Snowflake conceptsSnowPro CoreArchitecture
Snowflake architecture for SnowPro Core: what to know
How SnowPro Core tests Snowflake's storage, compute and cloud services layers: what runs where, what gets billed, and which edition a feature needs.
SHOW commands and cloning use only cloud services compute. Yet cloud services are billed only when daily use exceeds 10% of daily warehouse usage.
01 What does SnowPro Core test about Snowflake's architecture?
A question names a command and asks what it costs, and you freeze between warehouse and cloud services. That is the pattern to expect: each feature belongs to the layer that runs it.
Where this domain sits in the exam
The whole exam
- 1.0Snowflake AI Data Cloud Features and Architecture31%
- 2.0Account Management and Data Governance20%
- 3.0Data Loading, Unloading, and Connectivity18%
- 4.0Performance Optimization, Querying, and Transformation21%
- 5.0Data Collaboration10%
Save the 'Where this domain sits in the exam' graphic as a PNG image PNG for LinkedIn or Pinterest
Storage keeps each table's data in micro-partitions. Compute is your virtual warehouses plus serverless features such as Snowpipe. Cloud services handles metadata work such as SHOW commands.
The three layers and what runs in each
Storage
- Table data in micro-partitions
- Clustering by data dimensions
Compute
- Virtual warehouses
- Serverless features such as Snowpipe
Cloud services
- SHOW commands and cloning
- INFORMATION_SCHEMA queries
- Listing files for COPY
Compute cost comes from warehouses, serverless features and cloud services.
02 Which work runs only in cloud services?
Once you can name the layers, the cloud services traps stand out. Cloning, SHOW commands and INFORMATION_SCHEMA queries use only cloud services. Poorly selective COPY commands raise that usage, because listing files runs there.
You pay only when daily cloud services use exceeds 10% of daily warehouse usage. METERING_DAILY_HISTORY shows the credits actually billed.
Cloud services are billed only above 10% of daily warehouse use.
03 What do warehouse questions check?
Next comes compute, where questions turn on behavior you can picture. Resizing a suspended warehouse provisions nothing until it next resumes. Suspending one shuts idle resources down at once, while busy ones finish their statements first.
A warehouse created with defaults is STANDARD and XSMALL, and without a GENERATION clause it defaults to Gen2 where the region offers it.
Resizing a suspended warehouse takes effect when it next resumes.
04 Which storage features depend on the edition?
Last, storage features carry edition rules worth one flashcard each. Automatic Clustering is available in Standard Edition, while search optimization and materialized views need Enterprise or higher. Unlike those two, Automatic Clustering reorganizes existing data instead of adding storage.
Before your next mock exam, sort every architecture question you miss into its layer, and see which layer keeps catching you.
Search optimization and materialized views need Enterprise Edition or higher.
Summary
- Storage keeps table data in micro-partitions.
- Cloning, SHOW and INFORMATION_SCHEMA queries use only cloud services.
- Cloud services are billed above 10% of daily warehouse usage.
- New warehouses default to STANDARD, XSMALL and, where available, Gen2.
Sources and editorial status
Research checked Nov 1, 2026.